You are on page 1of 6

A minimal system allowing tubulation with molecular

motors pulling on giant liposomes


Aurélien Roux*†, Giovanni Cappello†, Jean Cartaud‡, Jacques Prost†, Bruno Goud*§¶, and Patricia Bassereau†§¶
*Laboratoire Mécanismes Moléculaires du Transport Intracellulaire, Unité Mixte de Recherche 144 Centre National de la Recherche Scientifique兾Institut
Curie, 26 Rue d’Ulm, 75248 Paris Cedex 05, France; †Laboratoire Physico-Chimie Curie, Unité Mixte de Recherche 168 Centre National de la Recherche
Scientifique兾Institut Curie, 26 Rue d’Ulm, 75248 Paris Cedex 05, France; and ‡Laboratoire de Biologie Cellulaire des Membranes, Institut Jacques Monod,
Unité Mixte de Recherche 7592 Centre National de la Recherche Scientifique兾Universités Paris 6 et 7, 2 Place Jussieu, 75251 Paris Cedex 05, France

Communicated by Pierre-Gilles de Gennes, Ecole Superiéure de Physique et Chimie Industrielles, Paris, France, February 22, 2002 (received for review
November 6, 2001)

The elucidation of physical and molecular mechanisms by which a proteins can directly bind to membranes by means of lipid
membrane tube is generated from a membrane reservoir is central binding domains. Whether these proteins are necessary for
to the understanding of the structure and dynamics of intracellular allowing motors to pull tubes in vivo is however still unclear.
organelles and of transport intermediates in eukaryotic cells. Membrane tubes can also be pulled out from membranes of
Compelling evidence exists that molecular motors of the dynein controlled lipid composition by hydrodynamic flow (19) and by
and kinesin families are involved in the tubulation of organelles. direct manipulation, using either micropipettes (20, 21) or
Here, we show that lipid giant unilamellar vesicles (GUVs), to which optical tweezers (22). These systems have been useful in under-
kinesin molecules have been attached by means of small polysty- standing the physics of membrane tube formation (19, 20, 23,
rene beads, give rise to membrane tubes and to complex tubular 24). However, the growth rates of the membrane tubes in these
networks when incubated in vitro with microtubules and ATP. systems are in the range of a few tens to a few hundred
Similar tubes and networks are obtained with GUVs made of micrometers per second, values that are significantly higher than
purified Golgi lipids, as well as with Golgi membranes. No tube the growth rates of tubes generated by motor proteins pulling on
formation was observed when kinesins were directly bound to the biological membranes in vivo [in the range of 1–2 ␮m兾s (5–8)].
GUV membrane, suggesting that it is critical to distribute the load This rate of in vivo membrane movement corresponds well to the
on both lipids and motors by means of beads. A kinetic analysis rates of in vitro motility of kinesin motors (the fastest—
shows that network growth occurs in two phases: a phase in which Neurospora kinesin—has a velocity value of about 2.6 ␮m兾s; see
membrane-bound beads move at the same velocity than free ref. 25).
beads, followed by a phase in which the tube growth rate de- The goal of this study was to determine whether binding a
creases and strongly fluctuates. Our work demonstrates that the motor protein to a lipid bilayer would be sufficient to generate
action of motors bound to a lipid bilayer is sufficient to generate membrane tubes. We show that a lipid reservoir, kinesin-coated
membrane tubes and opens the way to well controlled experi- beads, microtubules, and ATP provide a minimal system for
ments aimed at the understanding of basic mechanisms in intra- generating tubular structures that resemble tubes observed in
cellular transport. vitro with complex biological membranes (9–11).

Materials and Methods


B iological membranes, such as the endoplasmic reticulum
(ER), the Golgi apparatus, and endosomes, form elaborate
and highly dynamic tubular networks (1, 2). Recently, micros-
Reagents. Egg phosphatidylcholine (EPC), cholesterol, and
N-biotinyl-dioleyl-phosphoethanolamine (Biot-DOPE) were
copy of living cells has illustrated that membrane tubes also purchased from Avanti Polar Lipids. ␤-BODIPY 530兾550
participate in transport events between cellular compartments. C5-hexadecanoyl phosphatidylcholine, ␤-BODIPY 581兾591 C5-
For instance, long-range tubular transport intermediates have hexadecanoyl phosphatidylcholine, cholesteryl BODIPY 542兾563
been observed between Golgi and ER and between Golgi and C11, and cholesteryl BODIPY Fl C12 were obtained from Molecular
the plasma membrane, challenging the classical notion of small Probes. All chemicals were purchased from Sigma Aldrich except
spherical vesicles as the sole transport intermediates (3–8). ATP, GTP, and adenosine 5⬘[␤,␥-imido]triphosphate (AMP-PNP),
The formation and movement of membrane tubes in animal which were purchased from Roche Molecular Biochemicals.
cells is thought to involve an interaction of the membranes with Streptavidin beads (100 nm) were purchased from Bangs Labora-
the cytoskeleton, especially the microtubule network (9). Nu- tories (Carmel, IN). Biotinylated hemagglutinin-kinesin (a gift of F.
merous experiments, in vitro and in vivo, suggest a direct role of Nédélec, European Molecular Biology Laboratory, Heidelberg)
microtubules in the formation of the endoplasmic reticulum and was purified as described (26). By dividing the number of kinesins
Golgi membrane networks (9–11). Microtubule-dependent tu- by the number of beads, the number of kinesin molecules per bead
bulation of Golgi and endosomal membranes is amplified after was estimated at approximately 1,500.
the treatment of cultured cells with the fungal metabolite
brefeldin A (BFA) (12, 13). The interactions of membranes with Giant Unilamellar Vesicles (GUVs). GUVs were prepared by the
microtubules are mediated by several classes of proteins, notably electroformation technique (27). Rat liver Golgi membranes
motor proteins of the dynein and kinesin families (14). The were purified according to a standard procedure (28). Golgi
involvement of these motors in the movement of various or- lipids were then prepared as described (29).
ganelles, including transport intermediates and cellular struc-
tures, along microtubules is now well established. Their role in
Abbreviations: GUV, giant unilamellar vesicle; EPC, egg phosphatidylcholine; IMI, imida-
the formation of membrane tubes is less clear, but it has been zole; DIC, differential interference contrast.
reported that BFA-induced tubulation of Golgi membranes §B.G. and P.B. contributed equally to this work.
requires both in vivo and in vitro microtubule-based motor ¶To whom reprint requests should be addressed. E-mail: bruno.goud@curie.fr or
activity (11, 15). It should be pointed out that several proteins patricia.bassereau@curie.fr.
involved in membrane fission events, such as endophilin, am- The publication costs of this article were defrayed in part by page charge payment. This
phiphysin, and the GTPase dynamin, have been shown to induce article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
the formation of tubules from liposomes in vitro (16–18). These §1734 solely to indicate this fact.

5394 –5399 兩 PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 www.pnas.org兾cgi兾doi兾10.1073兾pnas.082107299


Assay for Tube Formation. Coverslips were washed for 5 min in chamber was filled with a buffer containing 1 mM ATP. Ap-
sulfochromic acid, rinsed 4–6 times in milliQ water, once in 95% proximately 10 min after injection, membrane tubes started to
ethanol, and stored at 4°C in ethanol. A coverslip was then dried form (Fig. 1 a and b). Once generated, tubes continued to grow
under a nitrogen flux and incubated for 1 to 2 min in a along the microtubules. Side branching events at the intersec-
poly(L-lysine) solution (0.01% wt兾vol). A flow chamber was tions of the underlying microtubule network were subsequently
built by intercalating two sheets of parafilm between this poly(L- observed (Fig. 1 a and b), leading to the formation of a network
lysine)-coated coverslip and a clean slide. The chamber was of membrane tubes (Fig. 2a) often extending over 50 ␮m from
finalized by heating the stack for a few seconds at 100–150°C on the GUVs. No tubes were observed in the absence of either
a heater plate. Its volume was approximately 25 ␮l. Diluted kinesin molecules or ATP (data not shown), indicating that tubes
microtubules in IMI buffer (50 mM imidazole, pH 6.7兾50 mM form through the action of kinesin. Fig. 1c illustrates a tube with
NaCl兾2 mM EGTA兾1 mM MgCl2) were injected into the two beads at its tip, with additional beads distributed along the
chamber and incubated for a few minutes. The chamber was tube. The beads are in fact required for tube formation, as
rinsed with 50 ␮l of IMI buffer and rinsed with 50 ␮l of 5 mg兾ml demonstrated in the following way. To allow direct binding of
casein diluted in IMI buffer. After a 15-min incubation, the kinesin to the membrane, we exploited the fact that kinesins are
chamber was rinsed with 50 ␮l of MB buffer (IMI buffer plus 1 not released from microtubules in the absence of ATP (Fig. 1d
mM ATP and 10 ␮M Taxol). Streptavidin beads (5 ␮l) mixed Left). Kinesins and fluorescent Cy3-labeled streptavidin were
with 2.5 ␮l of 5 mg兾ml casein were sonicated for 15 min on ice. sequentially injected into the chamber in the absence of ATP,
IMI buffer (30 ␮l) was added to the beads, and 1 ␮l of this leading to the formation of streptavidin兾kinesin complexes
mixture was mixed and incubated for a few minutes with 5 ␮l of bound to microtubules. After injection of GUVs and ATP,
1 ␮M biotinylated kinesins. Then, 44 ␮l of MB buffer was added kinesins detached from microtubules and bound to the lipid
to the kinesin-coated beads, and the chamber was filled with this bilayer, as demonstrated by the transfer of fluorescence from
mixture. GUVs (1–5 ␮l) were injected into the chamber and microtubules to the GUVs (Fig. 1d Center). Under these con-
allowed to sediment on the coated coverslip. ditions, no tubes were observed up to 1 h after injection (Fig. 1d
A similar assay was used for biotinylated Golgi membranes Right). By reducing 100-fold the amount of biotinylated lipids

CELL BIOLOGY
except that a smaller chamber volume (12 ␮l) and twice as many incorporated in the bilayer, we were able to show that the
kinesins per bead were used for pulling out tubes. absence of tube formation was not because of a high membrane
rigidity resulting from dense streptavidin grafting. On the other
Tube Imaging. Tubes were visualized by fluorescence confocal hand, tubes could be formed from these GUVs when kinesin-
microscopy, differential interference contrast (DIC) microscopy, coated beads were added (data not shown).
and reflection interference contrast microscopy (RICM) (30). Membrane tubes can be readily observed by confocal fluo-
The fluorescence intensity profiles of tubes were measured on rescence microscopy if fluorescent lipids are incorporated into
digital confocal images. The maximum was taken as the value of the vesicles. As shown in Fig. 2 a and b, the membrane tubes
the fluorescence intensity of one tube. To avoid any saturation showed a broad distribution in fluorescence intensity. The
or nonlinear effect, highly fluorescent tubes were omitted. On histogram of the normalized intensities is discontinuous, and
each network of tubes, all intensity values were normalized by only multiples of a unitary value are seen (Fig. 2d). Such a
dividing by the smallest intensity value. This operation was distribution implies that some tubes are either multilamellar or
repeated on many images to obtain an accurate estimation of the composed of bundles of unitary tubes. To define the precise
fluorescence distribution. structure of the network, we analyzed it by transmission electron
microscopy (Fig. 2 e–g). Single tubes had a constant diameter
Electron Microscopy. We built a chamber made of four 300-␮m (40 ⫾ 10 nm), a value close to that estimated for membrane tubes
mesh grids aligned in the middle of a coverslip, covered with a in vivo. Interestingly, two or more tubes were frequently ob-
collodion film, and dried at 60°C. The coverslip was incubated served aligned along a single microtubule (Fig. 2g), suggesting
for 1 min with poly(L-lysine), washed in water, and dried under that tubes of high fluorescent intensity (see Fig. 2 a and b) in fact
a nitrogen flux. A slide was then prepared with 2 spacers made represent bundles of several tubes. This hypothesis is supported
of 2 parafilm layers each and fixed by melting with a heater. After by the fact that the tubes with the highest fluorescence intensity
solidification of the spacers, the coverslip was assembled on often split into several tubes of lower intensity (Fig. 2b). In
them, keeping the grids aligned between the spacers and fixed addition, growth or retraction of a tube along another tube was
with vacuum grease. The assay was then performed as described frequently observed (Fig. 2c).
above. To avoid nonspecific adsorption onto the glass, 10 ␮g兾ml Based on growth velocity measurements, we were able to
of bacitracin in IMI buffer was used instead of casein. Negative distinguish two phases in the formation of a tubular network. At
staining was achieved by rinsing the chamber with 100 ␮l of 1% the first phase (I), the average growth velocity was 340 ⫾ 40
(wt兾vol) uranyl acetate in water and drying with a Whatman nm兾s, a value close to that measured for kinesin-coated beads
filter paper. The chamber was opened, and the grids were moving along microtubules in the absence of membranes (Fig.
removed from the coverslip. Carbon was then evaporated onto 3a). During this phase, the length of a tube increases essentially
the grids. Grids were observed in a Philips (Eindhoven, The linearly with time (Fig. 3c). Phase I corresponds to the time
Netherlands) EM 12 electron microscope fitted with a LaB 6 following the emergence of tubes from a given vesicle, up to the
filament and operating at 80 kV. Micrographs were taken on beginning of network formation (see Fig. 1a). Network growth
Kodak electron microscope films. then enters gradually in a second phase (phase II) in which the
average velocity decreased by a factor of two as compared with
Results phase I (Fig. 3g, EPC I and II). The observation of single tubes
Biotinylated kinesins (consisting of the motor domain of Dro- shows a slowing of growth, and even a complete stop, after a
sophila melanogaster kinesin) were attached to the membrane of growth period at constant velocity (data not shown). Moreover,
biotinylated GUVs (diameter ranging from 5 to 50 ␮m) by means the instantaneous velocity of the tip of the tube (Fig. 3f )
of 100-nm polystyrene beads coated with streptavidin. The assay fluctuated to a much greater extent than in phase I (Fig. 3d),
was first developed by using lipid bilayers consisting of 95% EPC where the fluctuations were similar to those of a kinesin-coated
and 5% biotinylated dioleyl-phosphoethanolamine (Biot- bead not bound to GUVs moving along a microtubule (Fig. 3b).
DOPE). GUVs and the kinesin-coated beads were injected into Phase II corresponds to a progressive densification of the
a chamber coated with Taxol-polymerized microtubules, and the network, and the number of growing tubes gradually decreased.

Roux et al. PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5395


Fig. 1. Formation of tubes and networks from EPC GUVs. (a) DIC images recorded with a charge-coupled device camera (time lapse, 12 s) showing several tubes
growing simultaneously from GUVs (bright spherical objects visible at the bottom of the image). Branching events occur on images 1, 4, and 5, leading to the
formation of a network (images 6 –9). (b) DIC images (time lapse, 7 s) showing a tube growing along a microtubule that splits into 2 tubes at the intersection
of 2 microtubules (image 7). A retraction event is visible on images 12 and 13, followed by further growth (images 14 –18). Black arrows point to microtubules.
(c) Tube growth observed by reflection interference contrast microscopy. Beads appear black or white depending on their distance to the substrate. Two beads
(gray arrow) are visible at the tip of a growing tube. Beads are also present along the tube (white arrow) growing on a microtubule (black arrow). (d) Tubes do
not form in the absence of beads. From Left to Right: without ATP, biotinylated kinesins-fluorescent Cy3 streptavidin complexes were fixed on the microtubule
network. In the presence of 1 mM ATP, the complexes were transferred onto the GUV (Center). Absence of tubes growing from a GUV as shown by DIC microscopy
(Right). [Bar ⫽ 5 ␮m.]

No significant, further growth activity was detected 30–60 min Finally, we wished to determine whether tubes could be
after the beginning of the experiment, which was not a result of obtained by binding the kinesin-coated beads to membrane
ATP exhaustion (data not shown). proteins instead of lipids. For this purpose, purified rat liver
We next tested the influence of the lipid composition of the Golgi membranes were incubated with N-hydroxysuccinimide-
membrane on the generation of tubes. The addition of choles- biotin, a reagent that biotinylates amino groups in proteins. The
terol (from 16 to 50% no.兾no., in the initial lipid solution) to the addition of kinesin-coated beads led to the formation of tube
EPC bilayer did not significantly change either the growth networks resembling those described above (Fig. 4). However, a
velocity (Fig. 3g, Chl) or the diameter of the fluorescent tubes phase I dynamic was not observable. Rather, the growth velocity
(data not shown). To obtain a more complex lipid composition, of the tubes resembled that of phase II in GUVs (Fig. 3g, Gol).
GUVs were made of lipids prepared from purified rat liver Golgi Bundles of tubes were also present in the preparations. In control
membranes (GPL, Golgi-purified lipid). Golgi lipids are com- experiments, tube formation was not observed when biotinylated
posed of 50% (no.兾no.) phosphatidylcholine, 20% phosphoeth- Golgi membranes were incubated in buffer alone.
anolamine, 6% phosphatidylserine, 12% phosphatidylinositol,
8% sphingomyelin, and cholesterol (lipid ratio 0.16) (29, 31). Discussion
Kinetic parameters comparable to the ones found with EPC The minimal system described above has two dynamic regimes.
GUVs were obtained, in particular for velocity values during When the growth velocity of tubes is nearly identical to the
phases I and II (Fig. 3g, GPL I and II). In addition, Golgi lipid velocity of a free bead (phase I in our assay), this implies that
GUVs also gave rise to networks consisting of bundles of the force experienced by individual motors is smaller than 1 pN
membrane tubes. (32). In our assay, the velocity is such that the force necessary

5396 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.082107299 Roux et al.


CELL BIOLOGY
Fig. 3. Kinetics of tube growth. Typical length (L) vs. time (t) and correspond-
ing instantaneous velocity (V) vs. time (t) plots, for, respectively: (a and b) a
free bead in a standard motility assay performed under the same conditions
(L corresponding to the distance covered by the bead); (c and d) an EPC tube
in phase I; (e and f ) an EPC tube in phase II. (g) Average velocities: free bead
(Bd), EPC lipid tubes in phases I and II (EPC), Golgi-purified lipid tubes (GPL) in
phases I and II, biotinylated purified Golgi membranes in phase II (Gol), and
EPC lipids with the addition of 50% (no.兾no.) cholesterol (Chl) in phase II.

Fig. 2. Tube networks. (a and b) Fluorescence confocal microscopy of EPC phospholipids and r ⬇ 20 nm from electron microscopy mea-
tubes containing 1% (no.兾no.) fluorescent BODIPY 530兾550 C5-HPC. A single surement, we obtain f ⬇ 12 pN. From geometrical arguments
representative confocal section is shown in a (the GUV appears out of focus). based on motor steric exclusion, bead curvature, and microtu-
(c) Sequence showing the growth of a fluorescent tube along another tube bule diameter (33), the number of motors likely to interact
(time lapse, 30 s). (d) Histogram of the normalized fluorescence intensities (NI), simultaneously with a microtubule at any given time can be
showing the discrete distribution of the fluorescence intensities of approxi- estimated to be between 10 to 20. Therefore, the load per motor,
mately 150 tubes. (e–g) Transmission electron microscopy images of EPC tubes
assuming that pulling is carried out by a single bead, is of the
(N); (e) Network of membrane tubes (white arrow) growing on a microtubule
network (black arrow). The GUVs are no longer visible, probably because of order of or smaller than 1 pN. After a few minutes, the growth
washing out during the staining process. ( f) A single tube aligned along a regime gradually reaches a second stage, in which tube growth
microtubule. The estimated diameter of this tube is 47 nm. (g) Bundles of 2 velocity is reduced by a factor of two. This observation indicates
tubes on a microtubule. [Bars ⫽ 5 ␮m (for a, b, and e) and 0.5 ␮m ( f and g)]. that each individual motor experiences a force about half stall
force (⬇2–3 pN) (32). Such an increase in force may result from
an increase in membrane tension, likely caused by membrane
for pulling a tube can be estimated from static arguments (20): extraction from GUVs (34). A high initial tension of the purified
f ⫽ ␲r␴ ⫽ 2␲␬兾r, in which ␴ is the membrane tension, and Golgi membrane could explain why phase I is not observed in
r ⫽ (␬兾2␴)1/2 is the tube radius (␬ is the membrane-bending this system. The velocity fluctuations in phase II (Fig. 3f ) could
rigidity modulus). Based on values of ␬ ⬇ 4 ⫻ 10⫺20 J for pure reflect either fluctuations in tension (35, 36) or fluctuations in

Roux et al. PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5397


Fig. 4. DIC microscopy images of the network extracted from biotinylated rat
liver Golgi stacks. [Bar ⫽ 5 ␮m.]

the number of motors working at a given time in a high-tension


condition.
A striking finding is that beads linking motor proteins to the
membrane bilayer seem to be a key element in our assay. Under
a load of a few pN, single phospholipid molecules are extracted
from a membrane in a few seconds (36), and motors detach from
the microtubule at even higher rates (37). On the other hand, it
takes more than a minute to pull out a 30-␮m-long tube (36).
Therefore, the role of beads could be to distribute the load Fig. 5. Possible mechanisms for the formation of bundles in the network. (a)
among a large number of lipids and motors, thus avoiding lipid Single tubes, originating from different points of the same vesicle and moving
extraction or motor detachment. It is also important to empha- along different microtubules, converge at a distance from the vesicle on a
size that the binding of several motors to a bead increases its single microtubule and form bundles. (b) Single tubes can also grow along the
processivity. How motors bind to biological membranes is still same microtubule and form bundles near the GUV membrane if weak inter-
poorly understood, but it has been reported that a kinesin actions exist between tubes and microtubules. The contact points (gray ar-
rows) of the two tubes with the microtubule should be distant enough to
(KIF13A) directly interacts with coat proteins (38), large cyto-
prevent tubes from coalescence.
solic complexes that are recruited onto membranes and play an
essential role in the formation of transport intermediates in cells
(39, 40). A tentative extrapolation of our results is that coat
protein complexes, by forming patches on the membranes, fulfil In conclusion, our work demonstrates that the action of motor
a role similar to that of the beads in our assay. Lipid subdomains, proteins is sufficient to produce membrane tubules from a lipid
such as detergent-resistant membranes (rafts), could also have bilayer. That tubes have been obtained with GUVs of complex
the same function, if motors can directly bind to them. Inter- lipid composition, as well as with purified Golgi membranes,
estingly, a kinesin (KIFC3) that associates with triton-insoluble indicates that this assay will be worthwhile for determining the
membranes has been recently characterized (41). pertinent biophysical parameters and for addressing the role of
Another interesting observation is the existence of bundles of lipid domains and protein coats in the events underlying the
tubes in the network, in apparent contradiction with elastic tubulation of biological membranes. It will be also interesting to
energy minimization, which predicts that tubes elongated from investigate how growing tubules can detach (fission event) from
the same GUV should coalesce once generated [I. Derényi, the membrane reservoir.
personal communication (21)]. One possible explanation is that
tubes originating from different points on the GUV membrane We thank F. Nédélec for kindly providing the kinesin plasmid; F.
form bundles because of the topology of the underlying micro- Jülicher, I. Derényi, and M. Dogterom for stimulating discussions; J.
tubules network, allowing bundles to form far enough from the Pécréaux and L. Legoff for their assistance in data processing; and B.
base of the vesicle (Fig. 5a). However, most of the bundles that Antonny, M. Bornens, D. Cassel, D. Chatenay, P. Chavrier, B. Hoflack,
we observed are localized very close to the GUV (see Fig. 2a). L. Johannes, R. McDermott, F. Nédélec, F. Perez, and T. Surrey for
In this case, bundles of tubes could be formed if tubes connect critical reading of the manuscript. This work was supported by grants
at different points to the same microtubule (Fig. 5b). Attractive from the Centre National de la Recherche Scientifique (Program
interactions between tubes and microtubules should then stabi- Physique et Chimie du Vivant), the Institut Curie, and the Association
lize the first points of contact, preventing tube coalescence. pour la Recherche Contre le Cancer.

1. Lee, C. & Chen, L. B. (1988) Cell 54, 37–46. 5. Polishchuk, R. S., Polishchuk, E. V., Marra, P., Saverio, A., Roberto, B., Luini,
2. Mollenhauer, H. H. & Morré, D. J. (1998) Histochem. Cell Biol. 109, A. & Mironov, A. A. (2000) J. Cell Biol. 148, 45–58.
533–543. 6. White, J., Johannes, L., Mallard, F., Girod, A., Grill, S., Reinsch, S., Keller, P.,
3. Hirschberg, K., Miller, C. M., Ellenberg, J., Presley, J. F., Siggia, E. D., Phair, Tzschaschel, B., Echard, A., Goud, B. & Stelzer, E. H. K. (1999) J. Cell Biol.
R. D. & Lippincott-Schwartz, J. (1998) J. Cell Biol. 143, 1485–1503. 147, 743–760.
4. Martinez-Menarguez, J. A., Geuze, H. J., Slot, J. W. & Klumperman, J. (1999) 7. Keller, P., Toomre, D., Diaz, E., White, J. & Simons, K. (2001) Nat. Cell Biol.
Cell 98, 81–90. 3, 140–149.

5398 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.082107299 Roux et al.


8. Stephens, D. J. & Pepperkok, R. (2001) J. Cell Sci. 114, 1053–1059. 26. Surrey, T., Elowitz, M. B., Wolf, P. E., Yang, F., Nedelec, F., Shokat, K. &
9. Allan, V. J. & Schroer, T. A. (1999) Curr. Opin. Cell Biol. 11, 476–482. Leibler, S. (1998) Proc. Natl. Acad. Sci. USA 95, 4293–4298.
10. Dabora, S. L. & Sheetz, M. P. (1988) Cell 54, 27–35. 27. Angelova, M. I., Soléau, S., Méléard, P., Faucon, J. F. & Bothorel, P. (1992)
11. Robertson, A. M. & Allan, V. J. (2000) Mol. Biol. Cell 11, 941–955. Prog. Colloid Polym. Sci. 89, 127–131.
12. Lippincott-Schwartz, J., Donaldson, J. G., Schweizer, A., Berger, E. G., Hauri, 28. Slusarewicz, P., Hui, N. & Warren, G. (1994) in Cell Biology: A Laboratory
H. P., Yuan, L. C. & Klausner, R. D. (1990) Cell 60, 821–836. Handbook (Academic, San Diego), Vol. 1, pp. 509–516.
13. Lippincott-Schwartz, J., Yuan, L., Tipper, C., Amherdt, M., Orci, L. & 29. Brugger, B., Sandhoff, R., Wegehingel, S., Gorgas, K., Malsam, J., Helms, J. B.,
Klausner, R. D. (1991) Cell 67, 601–616. Lehmann, W. D., Nickel, W. & Wieland, F. T. (2000) J. Cell Biol. 151,
14. Hirokawa, N. (1998) Science 279, 519–526. 507–518.
15. Lippincott-Schwartz, J., Cole, N. B., Marotta, A., Conrad, P. A. & Bloom, G. S. 30. Rädler, J. & Sackmann, E. (1993) J. Phys. II [French] 3, 727–748.
(1995) J. Cell Biol. 128, 293–308. 31. van Meer, G. (1998) Trends Cell Biol. 8, 29–33.
16. Farsad, K., Ringstad, N., Takei, K., Floyd, S. R., Rose, K. & De Camilli, P. 32. Visscher, K., Schnitzer, M. J. & Block, S. M. (1999) Nature (London) 400,
(2001) J. Cell Biol. 155, 193–200. 184–189.
17. Takei, K., McPherson, P. S., Schmid, S. L. & De Camilli, P. (1995) Nature 33. Badoual, M. (2001) Ph.D. thesis (Université Pierre et Marie Curie, Paris).
(London) 374, 186–190. 34. Helfrich, W. & Servuss, R.-M. (1984) Nuovo Cimento 3 D, 137–161.
18. Takei, K., Haucke, V., Slepnev, V., Farsad, K., Salazar, M., Chen, H. & De 35. Sandre, O., Moreaux, L. & Brochard-Wyart, F. (1999) Proc. Natl. Acad. Sci.
Camilli, P. (1998) Cell 94, 131–141. USA 96, 10591–10596.
19. Waugh, R. E. (1982) Biophys. J. 38, 19–37. 36. Evans, E. & Ludwig, F. (2000) J. Phys. Condens. Matter 12, 315–320.
20. Evans, E. & Yeung, A. (1994) Chem. Phys. Lipids 73, 39–56. 37. Coppin, C. M., Pierce, D. W., Hsu, L. & Vale, R. D. (1997) Proc. Natl. Acad.
21. Evans, E., Bowman, H., Leung, A., Needham, D. & Tirrell, D. (1996) Science Sci. USA 94, 8539–8544.
173, 933–935. 38. Nakagawa, T., Setou, M., Seog, D., Ogasawara, K., Dohmae, N., Takio, K. &
22. Raucher, D. & Sheetz, M. P. (1999) Biophys. J. 77, 1992–2002. Hirokawa, N. (2000) Cell 103, 569–581.
23. Bukman, D. J., Yao, J. H. & Wortis, M. (1996) Phys. Rev. E Stat. Phys. Plasmas 39. Schekman, R. & Orci, L. (1996) Science 271, 1526–1533.
Fluids Relat. Interdiscip. Top. 54, 5463–5468. 40. Wieland, F. & Harter, C. (1999) Curr. Opin. Cell Biol. 11, 440–446.
24. Heinrich, V., Bozic, B., Svetina, S. & Zeks, B. (1999) Biophys. J. 76, 2056–2071. 41. Noda, Y., Okada, Y., Saito, N., Setou, M., Xu, Y., Zhang, Z. & Hirokawa, N.
25. Steinberg, G. & Schliwa, M. (1996) J. Biol. Chem. 271, 7516–7521. (2001) J. Cell Biol. 155, 77–88.

CELL BIOLOGY

Roux et al. PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5399